Previous research has clearly implicated problematic attributional style, negative cognitive errors, and poor self-concept as risk factors for depression. A growing number of studies have tested cognitive models of depression in children. Relatively little theoretical or empirical work has examined the developmental origins of such cognitive risk factors. The overarching purpose of the current project is to take the next step backwards along this etiological chain by investigating the development of these putative cognitive diatheses in childhood. The first specific goal is to examine the structure of depressotypic cognitions at specific points during middle childhood (i.e., grades, K, 2, 4, & 6) in a large-scale, community-based, cross-sectional investigation. The hypothesis is that a few relatively global factors will differentiate into the depressive cognitive factors that characterize depression in adults. The second specific goal is to test several possible predictors of depressive cognitions and to map the developmental trajectories of these cognitions over the course of middle childhood and early adolescence. These tests will derive from a 4-year, longitudinal investigation that follows the initial cross-sectional study participants. The hypothesis is that problematic parenting style, aversive and uncontrollable negative events, and negative feedback from significant others will predict the emergence of depressive cognitive style. The third goal is to examine the incremental validity and the specificity of these cognitive factors vis-a-vis the prediction of depressive symptoms and first onset depressive disorder. We will address this goal in the context of a high-risk design that is nested within the larger longitudinal investigation. The hypothesis is that support for specific diathesis-stress models will become stronger as a function of the child's cognitive developmental level. Measurement strategies include one-on-one clinical diagnostic interviews with children and parents, behavioral observations of parent-child interactions, teacher observations, peer nominations, and self-reports. Methods for testing these hypotheses include multigroup confirmatory factor analysis, hierarchical linear modeling, cross-domain latent growth curve analysis, and structural equation modeling -- all with restricted maximum likelihood estimation or random-effects pattern-mixture methods for handling missing data. Findings from these investigations will (a) complement current cognitive theories of depression that here-to-fore have only vaguely alluded to the developmental origins of depressogenic cognitions and (b) provide a basis for developmentally informed prevention and early intervention programs.